AEC TECH^ |
38 |BUILDING DESIGN+CONSTRUCTION | October 2019how each of its departments uses space, and
to project future capacity needs. ZGF surveyed
the workers who would be moving, to gauge
more precisely how they worked individually and
with each other. Sara Howell and Amy Triscott,
ZGF’s Project Architect and Associate Urban
Designer, say the surveys revealed, for instance,
that two- to three-person meeting rooms were
in greater demand than expected. “One of our
goals was to right-size the ratio of different
spaces,” says Triscott.
SmithGroup’s Chicago offi ce, with about 100
people on one fl oor, was its fi rst to go com-
pletely agile. As the fi rm was fi tting out the fi rst
three-quarters of that space, it tagged employ-
ees and used a Bluetooth beacon to monitor
their movements. White says this data informed
how the last quarter of the fl oor was fi tted out,
and showed a need for smaller conference
rooms. (They also showed that people don’t
move around that much.) SmithGroup has also
created an app called Colleague Finder, which
uses access points, like smartphones, to locate
where people are in its offi ces. This was fi rst
tested at the fi rm’s Ann Arbor, Mich., offi ce and
is being rolled out to San Francisco (with 170
employees on several fl oors) and Detroit (300
people on three fl oors).IS DATA A RELIABLE CRYSTAL BALL?
The dream of many AEC fi rms is that data will
be their ticket to predicting outcomes, which
could, among other things, mitigate jobsite risk,
improve occupant comfort and, on a broader
scale, facilitate smarter cities.
To that end, Suffolk Construction is one of
nine construction fi rms that are members of the
Predictive Analytics Strategic Council, whosegoal, says Chin, is to share and aggregate
data for the purpose of developing predictive
software that can be marketed to the industry.
The impetus behind the council’s formation
was a 12-month collaboration between Suffolk
and Smartvid.io. Suffolk contributed a decade’s
worth of photo and project data, which Smart-
vid.io (the council’s technical advisor) analyzed
and then fed additional project data into a
multilayered machine learning model to see if
jobsite incidents could be forecast.
Chin and other AEC sources are convinced
that, with enough data, reliable predictions
are possible. “It’s a feasible aspiration,” says
Clark Construction’s Krause. But, he cautions,
the uncertain nature of construction will always
need to be built into any model. That’s why
he prefers prescriptive algorithms that leave
wriggle room for improvisation when it comes to
constructability, coding, and interoperability.
Nancy Reyes, HMC Architects’ Associate
Principal and Corporate BIM Director, says her
fi rm is actually less interested in predicting
behaviors than in leveraging data “that will give
us more options” to select from.
“I’m heartened that other fi rms aren’t being
deterministic,” says Gensler’s Tyson. The build-
ing environment, he points out, remains “highly
dynamic,” so what data provides “is an oppor-
tunity to learn and have information in real, or
almost real, time.”FROM RESPONSE TO REVENUE
Right now, construction data analysis and ap-
plication are part of what Jacobs’ McElvaney
and other AEC experts view as a “convergence
of technology” that is also spurring the rise of
digital twin, machine learning, and AI.
But just how robustly can data be monetized,
which some sources suggest is the industry’s
logical next step, and what the Predictive Analyt-
ics Strategic Council is trying to achieve?
Arup’s Kostura sees opportunities in data
for creating new services. He cites a large
mixed-use project in the U.S. where Arup fl owed
the buildings’ communication data through a
converged network in the cloud. He explains
that putting data in a single space allows for
the correlation of data sets, which is the basis
for machine learning. Creating a serverless built
environment, he adds, “enhances the building’s
operations.”+CLARK
CO
NSTR
UCTI
ON
Before starting most proj-
ects, Clark Construction col-
lects data on the location’s
underground utilities. That
investigation sometimes in-
cludes historical information,
such as previous surveys,
that is not always reliable or
correct, which requires Clark
to generate its own data
about the site to assess risk.
The image shown is from a
laser scan.